Audio transcription with SOME #911
Conversation
|
@maiko3tattun May I ask if the loading dialog can be used here. |
Perhaps it could be used. |
|
@oxygen-dioxide Maybe we should add checks to prevent very long chunks after slicing, for example, if someone clicks "transcribe" on an accompaniment audio that has no silence parts at all. The model contains attention operators and can cause OOM on very long sequences. |
OpenUtau/Views/MainWindow.axaml.cs
Outdated
| //msgbox?.SetText($"{text} {part.name}\n{wavPosS}/{wavDurS}"); | ||
| msgbox.SetText(string.Format("{0} {1}\n{2} / {3}",text, part.name, wavPosS, wavDurS)); | ||
| }); | ||
| transcribeTask.Wait(); |
There was a problem hiding this comment.
Instead of Wait() use ContinueWith(). See RegenFrq() for example. Modify project and show/hide dialog on scheduler from TaskScheduler.FromCurrentSynchronizationContext().
OpenUtau/Strings/Strings.axaml
Outdated
| @@ -8,6 +8,8 @@ | |||
| <system:String x:Key="context.part.delete">Delete part</system:String> | |||
| <system:String x:Key="context.part.rename">Rename part</system:String> | |||
| <system:String x:Key="context.part.replaceaudio">Reselect audio file</system:String> | |||
| <system:String x:Key="context.part.transcribe">Transcribe</system:String> | |||
There was a problem hiding this comment.
Even though transcribe is correct here, a more specific description like "Transcribe audio to create a note part" could be helpful.
oxygen-dioxide
changed the title
This PR adds integration with SOME, an audio to musical notes converter. Users can right-click on a wave part and select "Transcribe" to convert it into a voice part.
Before using, please download the SOME model from https://github.com/xunmengshe/OpenUtau/releases/download/0.0.0.0/some-0.0.1.oudep